Abstract
Cytoskeletal rearrangement during axon growth is mediated by guidance receptors and their ligands which act either as repellent, attractant or both. Regulation of the actin cytoskeleton is disturbed in Spinal Muscular Atrophy (SMA), a devastating neurodegenerative disease affecting mainly motoneurons, but receptor-ligand interactions leading to the dysregulation causing SMA are poorly understood. In this study, we analysed the role of the guidance receptor PlexinD1 in SMA pathogenesis. We showed that PlexinD1 is cleaved by metalloproteases in SMA and that this cleavage switches its function from an attractant to repellent. Moreover, we found that the PlexinD1 cleavage product binds to actin rods, pathological aggregate-like structures which had so far been described for age-related neurodegenerative diseases. Our data suggest a novel disease mechanism for SMA involving formation of actin rods as a molecular sink for a cleaved PlexinD1 fragment leading to dysregulation of receptor signaling.
Introduction
Spinal muscular atrophy (SMA, OMIM #253300) is a progressive neurodegenerative disease which affects newborns in its most severe and frequent form. SMA displays an incidence of 1: 6,000 - 10,000 (1,2). Patients suffer from proximal muscle weakness and atrophy due to degeneration of α-motoneurons in the spinal cord (3). SMA is caused by functional loss of the ubiquitously expressed survival of motoneuron 1 (SMN) protein due to deletion, gene conversion or mutation in the SMN1 gene (4,5).
There is growing body of evidence of perturbations in actin cytoskeletal regulation in SMA. It has been hypothesized that these mechanisms result in altered stability of neuromuscular junctions (NMJs), muscle denervation and atrophy (6). We and others previously described mechanisms of impaired F-actin dynamics, in growth cones comprising SMN and its functional interaction with RhoA-associated coiled-coil kinase (ROCK) (7–12). In line with this, dysregulation of the actin-binding proteins cofilin and profilin acting downstream of ROCK and disturbed neurite outgrowth have been observed (7–13). Nonetheless, upstream mechanisms altering actin cytoskeletal regulation are still elusive.
PlexinD1 (PLXND1) is a membrane-spanning cell surface guidance receptor for secreted class 3 Semaphorins (SEMAs) and is highly expressed in the central nervous system, endothelial cells and cells of the immune system (14). The receptor comprises an intracellular segmented GTPase-activating (-GAP) domain interconnected by a Rho-GTPase-binding domain (RBD) (15). Therefore, the receptor can modulate actin cytoskeletal dynamics in the growth cone via small GTPases. Normally, it acts as a repulsive receptor after ligand binding, but depending on co-receptor composition, it can also mediate attraction (16,17). Thus, its function in axonal guidance and growth cone stability made PLXND1 an interesting target to be analysed with respect to SMA.
In this study, we show that PLXND1 is intracellularly cleaved by metalloproteases (MPs) in cellular and mouse models for SMA. Cleavage thereby switches the receptor’s response to its ligand SEMA3E from attraction to repulsion which can be fully reversed by MP inhibition. Moreover, the cleaved PLXND1 fragment binds to elongated structures within different cellular compartments, which we were also able to reduce by MP inhibition. Additionally, we have been able to identify the elongated structures as actin rods. These are pathological actin-containing structures known from other neurodegenerative diseases, which are here first described for SMA. In this study, we elucidated a functional link between PLXND1 cleavage and rod formation pointing towards putative novel treatment strategies in SMA.
Results
PLXND1 is mislocalised in SMN-depleted motoneuron-like cells
As a first step to elucidate the role of PLXND1 in SMA, we analysed Plxnd1 expression in a motoneuron-like NSC34 cell culture model for SMA which has been established previously (11,18). NSC34 cells were transfected with siRNA against murine Smn (siSmn) or scrambled control siRNA (siCtrl) and then differentiated for three days. Subsequently, mRNA was isolated and relative Plxnd1 expression was analysed by qRT-PCR revealing that Plxnd1 was less expressed in SMN-depleted cells compared to control cells (Fig. 1A). Western blot analysis instead showed that PLXND1 protein level was unchanged indicating that it was stable upon SMN knockdown (Fig. 1B and C). Antibody specificity was evaluated upon overexpression or knockdown of PLXND1 (Supplementary Material, Fig. S1A). However, we found a remarkably altered localisation of the receptor (Fig. 1D and Supplementary Material, Fig. S1B). In control cells, PLXND1 was distributed within the membrane, while SMN-depleted cells additionally harboured elongated structures in the nucleus and, to a lesser extent, in the cytoplasm. It could thus be suggested that the decrease of the Plxnd1 transcripts serves as a feedback mechanism to prevent excessive PLXND1 formation. The formation of those aggregates increased over time of SMN knockdown (Fig. 1E). About 7% of control and SMN knockdown cells showed PLXND1-positive aggregates 24 h post transfection. While the number of those structures remained stable over time in control cells, the number increased to 23% 48 h post transfection and to 50% 72 h post transfection under SMA conditions. To rule out the possibility of unspecific staining, different permeabilisation methods were tested, but elongated PLXND1-structures were visible under all conditions (Supplementary Material, Fig. S1C). As an additional control, we used another antibody detecting PLXND1 and also detected the tubular structures (Supplementary Material, Fig. S1D). To test whether other Plexins form these structures, PLXND1-positive structures were analysed for PLXNB1 co-staining as a representative of another sub-family of Plexins. PLXNB1 did not co-localise with PLXND1 (Supplementary Material, Fig. S1D) indicating a specific localization pattern of PLXND1 in SMA but not in control cells. Together, these data show that NSC34 cells express Plxnd1 and that there is a dysregulation upon SMN knockdown.
PLXND1 mislocalises in SMN knockdown cells. NSC34 cells were transfected with scrambled control RNA (siCtrl) or siRNA against Smn (siSmn) and differentiated for three days. (A) Gene expression determined by qRT-PCR (mean ± SEM, n = 5, paired two-tailed t-test, **P < 0.01). (B) Representative Western blot probed for PLXND1 (ab96313) and tubulin as housekeeping (uncropped blots Supplementary Material, Fig. S8A-B). (C) Quantification of PLXND1 signal compared to tubulin (mean ± SEM, n = 6, paired two-tailed t-test, n.s.). (D) Representative immunocytochemical staining of PLXND1 (ab28762) with DAPI as nuclear counterstaining. Asterisks indicate elongated structures. Scale bar 20 µm. (E) Quantification of PLXND1-positive structures depending on SMN knockdown (mean ± SEM, n = 3, one-way ANOVA with Sidak post-test, **P < 0,001, ****P < 0.0001). Uncropped blot in Supplementary Material, Figure S8C.
PLXND1 binds to elongated structures in spinal cord from SMA mice
To confirm the in vitro results we employed a severe SMA mouse model (FVB.Cg-Tg(SMN2)2Hung Smn1tm1Hung/J (19), denominated as Taiwanese SMA mice). Previous studies have shown by Northern blot and in-situ hybridization that Plxnd1 is highly expressed during embryogenesis in mouse and chicken but not in adult murine and human CNS (20,21). Thus, we first analysed relative Plxnd1 expression in total spinal cord of mutant mice (Smn−/−;SMN2tg/0) and their healthy control littermates (Smn+/−;SMN2tg/0) at asymptomatic (P1), early symptomatic (P5) and symptomatic (P8) stages by qRT-PCR (Fig. 2A). Conversely to other studies, we here detected Plxnd1 transcripts in control animals at any developmental stage evaluated. Interestingly, Plxnd1 expression was increased in mutant mice at P1 and P5 and significantly decreased at P8 reflecting the results obtained in NSC34 cells (Fig. 1A). Furthermore, no differences of PLXND1 protein levels between control and mutant animals were observed similar to the SMA model in NSC34 cells (Fig. 2B and C). Next, we stained lumbar spinal cord sections from P0, P5 and P8 mutant and control mice for PLXND1. In these sections, motoneurons were identified by their morphology and by immunostaining of neurofilament H with an SMI32 antibody (Fig. 2D). PLXND1 mainly localised to SMI32-positive motoneurons, but also to other cells of the spinal cord, presumably endothelial cells. Quantification of PLXND1 immunofluorescence of SMI32-positive cells showed no significant difference between control and mutant mice at P0, P5 or P8 (Fig. 2E). To rule out the possibility of unspecific staining, the primary antibody was omitted and only low background but no distinct staining was detected (Fig. 2D). Moreover, four additional antibodies detecting different epitopes of the protein were included to stain PLXND1 (Supplementary Material, Fig. S2A and B). Though background was higher using the other antibodies, all of them showed the same PLXND1 distribution in the spinal cord as observed above. Our data therefore clearly indicate that Plxnd1 is expressed in spinal cord using qRT-PCR and immunofluorescence and that protein levels are unchanged in mutant mice compared to controls.
PLXND1 expression in spinal cord of control and mutant mice. (A) Gene expression in total spinal cord from non-symptomatic (P0), early symptomatic (P5) and symptomatic (P8) control (Smn+/−;SMN2tg/0) and mutant (Smn−/−;SMN2tg/0) mice determined by qRT-PCR (mean ± SEM, n = 4, unpaired two-tailed t-test between genotypes within one timepoint, *P < 0.05). (B) Representative Western blot probed for PLXND1 (ab96313) and tubulin as housekeeping (uncropped blots Supplementary Material, Fig. S8D-E). (C) Quantification of PLXND1 signal compared to tubulin (mean ± SEM, n = 4 for P1, n = 5 for P5, n = 6 for P8, unpaired t-test between genotypes within one timepoint, n.s.). (D) Immunocytochemical staining of PLXND1 (ab28762) and SMI32 in the anterior horn of lumbar spinal cord. Scale bar 50 µm. (E) Quantification of PLXND1 immunolabelling in SMI32-positive cells (mean ± SEM, n as indicated, unpaired two-tailed t-test between genotypes within one timepoint, ns, non-signifcant).
Interestingly, further analysis revealed that SMI32-positive motoneurons and other cells of the lumbar spinal cord (sections between L1-2), which are not GFAP-positive astrocytes, showed PLXND1-positive elongated intracellular structures at P3, P5 and P8 but not at P0 (Fig. 3 and Supplementary Material, Fig. S3) indicating that the formation occurs prior to the onset of symptoms. Strikingly, also control animals harboured few PLXND1-positive elongated structures, but we were unable to reliably quantify them. However, subtle changes in SMN levels may result in the formation of those structures (Fig. 1E). This hypothesis is further supported by the fact that phenotypically unobtrusive control animals are heterozygous for Smn.
PLXND1-positive extended structures were found in the spinal cord of mutant mice. PLXND1-positive structures (ab28762) were found in lumbar spinal cord of P3, P5 and P8 mice, but not at the asymptomatic P0 stage (asterisks and insets). SMI32 was used for motoneuron staining. Scale bar 50 µm.
As the Taiwanese SMA mouse model reflects severe SMA, we wondered whether the elongated structures were also present in symptomatic mice from an SMN antisense oligonucleotide (ASO)-induced intermediate SMA mouse model (22). Therefore, Taiwanese mice were back-crossed on a C57BL/6 N background (23) and additionally treated at P2 and P3 with suboptimal doses of an ASO (22) increasing SMN protein level from the human SMN2 transgene (24) and using the same pharmacological mechanism as the just recently FDA-approved drug Spinraza (25–27). These intermediate mutant mice survive approximately 26 days (22). Spinal cord sections from P21 control and symptomatic intermediate mutant mice were stained for PLXND1 as above. Similarly, the elongated, PLXND1-positive structures were also present in SMI32-positive motoneurons in this SMA mouse model (Supplementary Material, Fig. S4) suggesting that SMN increase alone may not prevent the formation of the PLXND1-positive structures.
The elongated complexes are actin rods decorated with PLXND1
Next, we characterized the elongated PLXND1-positive structures. Cytoplasmic rod-like structures have been reported for other neurodegenerative diseases (28). Moreover, intranuclear rod-like structures are typical in a subgroup of congenital myopathies (29). Both types of rods contain large amounts of actin and are therefore denoted as actin rods. Transient rods are neuroprotective by preventing actin-mediated ATP decline during stress (30). Under permanent stress, however, these rods become persistent and lead to neurodegeneration (30–33).
Actin rods are defined by distinct properties: (i) Actin rods, which have been induced due to ATP depletion, comprise cofilin-saturated F-actin (28). (ii) Due to actin saturation by the actin-binding protein cofilin, phalloidin cannot compete for the F-actin binding site rendering rods phalloidin-negative (34). (iii) Rod formation is likely mediated by reactive oxygen species. These radicals lead to the formation of intermolecular disulphide bonds between unphosphorylated cofilin monomers, which then incorporate into actin rods. Thus, rods are susceptible to reductive agents, e.g. β-mercaptoethanol (β-ME) (34,35). Importantly, the elongated complexes observed in our study showed all of these properties in the cellular SMA model: SMN knockdown cells were stained for PLXND1, cofilin and actin. Confocal microscopy revealed that PLXND1-positive structures highly co-localised with both cytoskeletal proteins as quantitatively determined by intensity correlation analysis with calculation of PDM (product of the differences of the mean (36)) values (Fig. 4A). Moreover, the structures were negative for staining with fluorophore-tagged phalloidin (Fig. 4B). To test the influence of the reducing agent β-mercaptoethanol (β-ME) on the stability of the elongated structures, we treated cells for 30 min prior to fixation and counted the number of PLXND1-positive structures, which were decreased by incubation with the reducing agent (Fig. 4C). However, also control cells displayed a small number of these structures. The results show that the number of actin rods in SMN-depleted cells is increased and that PLXND1 is present in these rods. Moreover, our data suggest that oxidative stress may trigger rod formation in SMA.
Elongated structures are PLXND1-positive actin rods. (A) Confocal images of SMN-depleted cells stained for PLXND1 (antibody ab28762), cofilin and actin. Co-localisation was analysed within the insets representing their corresponding PDM (product of the differences of the mean) values. Positive values indicate co-localisation. Scale bar 10 µm. (B) SMN knockdown cells were stained for PLXND1 (antibody ab28762) and F-actin using Alexa546-tagged phalloidin as well as DAPI for nuclear counterstaining. Scale bar 50 µm. (C) Control and knockdown cells were treated with water or β-ME 30 min prior to fixation. The percentage of cells with PLXND1-positive rods was counted (mean ± SEM, n = 3, not repeated measures two-way ANOVA with Sidak post-test, ****P < 0.0001; paired two-tailed t-test, #P < 0.05).
PLXND1 is proteolytically cleaved in SMA and binds to actin rods
To further investigate PLXND1 binding, we enriched actin rods from knockdown cells using density gradient centrifugation (28) (Fig. 5A). The fractions were subsequently analysed by Western blot. Typically, cofilin and actin intensities decreased within the first fractions, but re-appeared in the fraction at the 10%/15% phase border in which actin rods should be enriched (28) (Fig. 5B). In addition, electron microscopy confirmed the presence of actin rods within that fraction. Negatively stained nearly 50 nm thick and about 1 µm long banded filaments stacked together in parallel and antiparallel direction to branched fibrils. Despite that, also up to 400 nm long helically arranged thinner filaments were concentrated in the same fraction (Fig. 5C). Thus, we further analysed the phase border-containing fraction for PLXND1 by Western blot (Fig. 5D). We detected the full-length PLXND1 represented by the 250 kDa band in the total cell lysate. Unexpectedly, we found decreased levels of full-length receptor and enrichment of a 70 kDa fragment in the rod-containing fraction. As the antibody used detects an epitope at PLXND1’s C-terminus, it is likely that this fragment corresponds to its cytoplasmic domain. To test this, we performed immunofluorescence analyses in SMN knockdown cells using an antibody detecting the extracellular N-terminal ectodomain (sc-46246) in comparison with an antibody detecting the C-terminal endodomain (ab96313) (Fig. 5E and Supplementary Material, Fig. S2A and B). Both antibodies displayed membrane localization confirming the presence of the full-length receptor at the cell surface. However, only the cytoplasmic domain-binding antibody was able to stain PLXND1-positive actin rods confirming indirectly the identity of the 70 kDa fragment detected in the rod fraction.
PLXND1 is cleaved by metalloproteases and sequestered to actin rods. (A) Schematic overview of actin rod enrichment (28): SMN knockdown cells were treated with nocodazole to disrupt interfering microtubules and lysed mechanically. After a pre-clarifying step, the supernatant containing rods was centrifuged on a two-phase OptiPrep gradient. Rods collected at the phase border. (B) Fractions from the gradient were analysed for the presence of actin and cofilin (uncropped blots Supplementary Material, Fig. S8F and G). Both signals decreased within the 10% phase but re-appeared at the phase border (fraction 7) where rods should collect. (C) Electron micrographs of indicated fractions revealed the presence of actin rods in fraction 7. Scale bar 100 nm. (D) Fraction 7 was further processed and the remaining pellet was analysed for PLXND1. Unexpectedly, a 70 kDa fragment of PLXND1 was detected within that fraction. (E) PLXND1 staining using antibodies detecting either PLXND1’s C-term (antibody ab96313) or N-term (antibody sc-46246). Overlap of the stainings showed that the antibody detecting the C-term but not the N-term stains PLXND1-positive actin rods. (F) PLXND1 reveals a putative cleavage consensus sequence (RXXR) close to the transmembrane domain. Cleavage would produce a 72 kDa fragment. (G) Control (siCtrl) and knockdown (siSmn) cells were treated with the metalloprotease inhibitor Batimastat (BB-94) during differentiation. Cells were fixed and the percentage of cells with PLXND1-positive rods was determined (mean ± SEM, n = 3, not repeated measures two-way ANOVA with Sidak post-test, ****P < 0.0001; paired two-tailed t-test, #P < 0.05).
Ectodomain shedding and intracellular cleavage of cell surface receptors is a common mechanism to regulate signal transduction and has been well-described for Plexins and other guidance receptors, e.g. ephrins and neogenin (37,38). Within the Plexin family, PLXNB1 is cleaved by metalloproteases (MPs) at the R1302RRR consensus sequence (39). This sequence is conserved in the intracellular domain of PLXND1 (R1299AER) in mouse and human (Genbank accession numbers NP_080652.2 and NP_055918.2). The predicted molecular weight of the cleavage product is 72 kDa which has the same size as the PLXND1 cleavage product found in the actin rod fraction (Fig. 5F). This prompted us to evaluate PLXND1 cleavage and tested the effect of the broad range MP inhibitor Batimastat (BB-94). BB-94 was added to the motoneuron-like NSC34 cells during differentiation and the number of PLXND1-positive actin rods was determined. As expected, this treatment led to a reduction of PLXND1-positive actin rods upon SMN knockdown, but had no effect in control cells (Fig. 5G). The slight but significant reduction indicates that the responsible MP may not be fully inhibited by BB-94. Taken together, these data show that PLXND1 is processed by MPs in SMA which can be inhibited by BB-94. Additionally, this cytoplasmic fragment binds to actin rods under SMA conditions.
PLXND1 cleavage switches its gating properties in SMA
SMN knockdown led to increased PLXND1 cleavage by MPs. Thus, we wondered whether this modification alters the function of the guidance receptor in SMN-depleted cells. Growth cones sense the extracellular environment and are either attracted or repelled by guidance cues. Thus, growth cone collapse is a valuable readout for the evaluation of the function of repulsive guidance receptors. Cells were therefore incubated with the PLXND1-specific ligand SEMA3E at 1 nM or BSA as a control for 30 min after which growth cone collapse was assessed by measuring phalloidin-Alexa546 positive growth cone areas (Fig. 6A and B). Interestingly, growth cones from SMN-depleted cells strongly collapsed. Conversely, growth cones from control cells had growth cone areas comparable to those observed for BSA treated cells indicating that PLXND1 is silent in this cell culture system or that it normally mediates attraction which cannot be measured in a growth cone collapse assay. To test this hypothesis, we performed the collapse assay in PLXND1-depleted cells (Supplementary Material, Fig. S5), in which knockdown efficiency was about 60% (Supplementary Material, Fig. S1A). Indeed, BSA had no influence on growth cone morphology whereas SEMA3E induced growth cone collapse. Together, these data show that PLXND1 cleavage results in a switch of SEMA3E-mediated signal gating properties.
SEMA3E induces growth cone collapse in SMN-depleted but not control cells. (A,B) SMN knockdown (siSmn) and control (siCtrl) cells were differentiated for three days and incubated with 1 nM BSA as control or SEMA3E for 30 min. Cells were stained with Alexa546-tagged phalloidin. Growth cone collapse (indicated by arrows) was assessed by measuring growth cone size (mean ± SEM, N = 100–200, n = 4, not repeated measures two-way ANOVA with Sidak post-test, *P < 0.05). (C,D) Cells were differentiated in the presence or absence of the metalloprotease inhibitor Batimastat (BB-94). SEMA3E treatment and growth cone collapse analysis was carried out as above (mean ± SEM, N = 100–200, n = 3, not repeated measures two-way ANOVA with Sidak post-test, **P < 0.01). Scale bar 20 µm.
As BB-94 treatment partially rescued PLXND1 cleavage (Fig. 5G), we wondered whether this compound is able to rescue growth cone collapse. Cells were differentiated in the presence or absence of BB-94 and growth cone area was assessed as above after SEMA3E incubation (Fig. 6C and D). Notably, treatment with BB-94 completely rescued the collapsing effect. These results indicate that processing of PLXND1 by MPs leads to altered SEMA3E response in SMN-depleted cells and that MP inhibition could rescue the defects.
PLXND1-decorated rods in iPSC-derived motoneurons of SMA patients
To link our observations to the situation in SMA patients, we analysed whether PLXND1-positive actin rods were also detectable in induced pluripotent stem cell (iPSC)-derived motoneurons from SMA patients and controls (Supplementary Material, Table S1). Similar to previous experiments (40), the average number of Isl-1-positive neurons was 20–25% and SMA iPSC motoneurons showed 80% reduced SMN expression compared to controls (Fig. 7A). Based on this evaluation, iPSC-derived control motoneurons were first cultivated for two weeks and stained for Tuj1 and PLXND1 (Supplementary Material, Fig. S6) showing broad, punctuate but not rod-shaped PLXND1 distribution. As expected, PLXND1 was highly enriched in growth cones. Afterwards, SMA motoneurons were stained for PLXND1 revealing elongated structures in the soma as well as in neurites, which were significantly increased compared to controls (Fig. 7B and C). Positive cofilin and negative phalloidin co-stainings confirmed that these structures were indeed PLXND1-positive rods (Fig. 7C and Supplementary Material, Fig. S7). Compared to rod numbers in SMN-depleted NSC34 cells, rod numbers were vastly lower in iPSC-derived motoneurons from SMA patients. This discrepancy could be due to different cellular environments and different cellular models. However, the results indicate that PLXND1 cleavage and rod formation is not restricted to severe and intermediate SMA mouse models, but may also have a significant pathogenic role in SMA patients.
PLXND1-positive rods in iPSC-derived motoneurons from SMA patients. iPSC-derived motoneurons from SMA patients or controls were cultivated for two weeks. (A) SMN levels in control and SMA iPSC-derived motoneurons (mean ± SEM, immunoblot performed in triplicates). (B,C) iPSC-derived motoneurons were stained for PLXND1 (ab28762) as well as cofilin and analysed by confocal microscopy. Scale bar 50 µm (Ctrl) and 20 µm (SMA). The percentage of cells with PLXND1-positive rods was counted for control and SMA iPSC-motoneurons (mean ± SEM, n = 3, unpaired two-tailed t-test, *P < 0.05).
Discussion
Growth cone dysfunction and axonal outgrowth defects in the motoneuron disease SMA have been hypothesised to result in impaired NMJ stability leading to muscle denervation and atrophy (6). It has been shown that ROCK, the master regulator of the actin cytoskeleton (41), is involved in these defects, but upstream mechanisms are poorly understood (7–12,18). Plexins are a large family of guidance receptors regulating the actin cytoskeleton in growth cones during guidance, synaptic specificity and maintenance (42). Moreover, they are involved in the development of the cardiovascular system (14).
In this study, we evaluated PLXND1 in SMA models and found novel properties of this receptor. PLXND1 is intracellularly cleaved by MPs generating a C-terminal fragment. This protease-mediated cleavage resulted in two different outcomes: First, cleavage alters PLXND1’s signalling response to SEMA3E. Second, the cytoplasmic fragment binds to actin rods in SMN-depleted NSC34 cells, in the spinal cord of (pre-)symptomatic mice of a severe as well as an intermediate mouse model and in motoneurons of SMA patient-derived iPS cells. Thus, hyperactivation of MPs in SMA has a dual effect on PLXND1 (Fig. 8).
Connection of SMN depletion, PLXND1 cleavage and actin rod formation: a mechanistic model. SMN depletion leads to increased MP activity which is responsible for PLXND1 cleavage. As a consequence, PLXND1‘s co-receptors signal to induce growth cone collapse which has not been observed in controls. Furthermore, RhoA-GTP is increased in SMA (7). Under stress, RhoA-GTP is cleaved into an N-terminal fragment, which displays binding property for ROCK, and into a C-terminal fragment, which has been shown to induce actin rods (58). Actin rods are stabilised by hyperphosphorylated cofilin which is oxidised during persistent stress (33). Actin rods serve as a sink for the PLXND1 cleavage product probably without inducing rods itself.
Members of the PlexinB family reveal an extracellular consensus sequence at which they undergo MP-mediated ectodomain shedding (39). Such an extracellular consensus sequence is not present in PLXND1, but a putative cleavage site has been found intracellularly close to the transmembrane domain (Fig. 8). Nonetheless, to our knowledge, PLXND1 cleavage has not been described yet. Here, we show that PLXND1’s specific ligand SEMA3E induces growth cone collapse in SMN-depleted NSC34 cells, but not in control cells. The receptor can assemble into a ternary complex consisting of PLXND1 and its co-receptors neuropilin-1 (NRP1) and vascular endothelial growth factor receptor-2 (VEGFR2) (16,17). Deletion of PLXND1’s cytoplasmic domain has no influence on morphological outcome indicating that VEGFR2 comprises the signal transducing element and that PLXND1 promotes SEMA3E binding to NRP1 in this tri-complex (16). Moreover, it is known for PLXNB1 that ectodomain shedding increases SEMA4D binding and signal propagation (39). This is in line with our results since PLXND1 cleavage did not prevent signalling. Instead, SEMA3E-mediated growth cone collapse was increased after PLXND1 cleavage or knockdown.
What could be the pathological outcome related to altered PLXND1 signalling? The specific PLXND1-SEMA3E interaction is important for fine-tuned development of several monosynaptic reflex arcs (42–44). Alterations in PLXND1-SEMA3E signalling in either proprioceptive sensory neurons or motoneurons lead to defects in monosynaptic connectivity and therefore to malformation of sensory-motor circuits. Interestingly, malfunctioning motor circuits have been described in SMA (45,46). It is likely that proprioceptive axons reach their target motoneurons until P0 but that specificity and maturation are disturbed during development (45). Moreover, reduced proprioceptive synapses on motoneurons in SMA mice were described which could be due to failure of sensory axon projection (specificity) or branching (maturation) (45). Importantly, we found in this study that PLXND1 cleavage is a postnatal event corresponding to the kinetics of circuit dysfunction in SMA. The PLXND1-SEMA3E interaction could provide the underlying molecular mechanism for this early deafferentiation characterizing SMA.
PLXND1 furthermore displays a relevant molecule during vascularisation. PLXND1 knockout mice are born with congenital heart disease and defects in aortic arch arteries (47) and cardiac phenotypes have also been reported in SMA patients (48). A recent study showed that the number of blood vessels is decreased in skeletal muscle and spinal cord of two severe SMA mouse models as well as type I SMA patients resulting in hypoxia and oxidative stress (49). Interestingly, Plxnd1 knockout mice show shorter tails compared to their control littermates (50). Shorter tails as well as tail necrosis followed by ear necrosis at later stages are phenotypic hallmarks of several mild and severe SMA mouse models (19,22,51,52). In line with this, peripheral necroses are also seen in some SMA type I patients (53). Thus, PLXND1 may display another link to SMA pathology by its influence on vascular development.
The second consequence of MP-mediated PLXND1 cleavage was that the cytoplasmic fragment binds to actin rods. We here report the presence of actin rods in different SMA models evaluated in different laboratories and SMA mouse models emphasising the reproducibility of this finding. Actin rods are pathological structures found in post mortem tissue from Alzheimer’s disease patients (31). Under stress, the formation of actin rods is initially neuroprotective: Transient oxidative stress leads to cofilin activation (i.e. hypo-phosphorylation) as well as ATP decline and rods remove free dephosphorylated cofilin as well as ADP-bound actin from the cell (30). Despite other mechanisms, this sequestration is thought to prevent stress-related cofilin shuttling into mitochondria and cytochrome release (31). However, constant oxidative stress and production of reactive oxygen species culminate in the formation of permanent rods by cross-linking the incorporated cofilin via disulphide bonds (35). These permanent rods block axonal transport and other physiological processes resulting in axonal dye-back mechanisms as shown in Alzheimer’s disease (30–33). The role of intranuclear rods is still controversial. Recent studies showed that actin rods alter chromatin structures, replace RNA polymerase II and disturb gene accessibility and transcription (54,55). Moreover, also Coronin binds to actin rods (56). Coronin is an actin-binding protein shown to rescue endocytosis defects in SMA (22). Sequestration of Coronin to the rods could impair its function during endocytosis seen in SMA (22,57). A main point is the mechanistic link between SMN and actin rod formation. Low SMN levels as in SMA lead to increased RhoA expression, cofilin hypophosphorylation and disturbed G-/F-actin dynamics due to dysregulated RhoA/ROCK signaling (7,10). Under stress, GTP-bound RhoA is proteolytically cleaved and the C-terminal RhoA fragment has been shown to induce phalloidin-positive intranuclear rods (58). ATP or oxidative stress triggers the formation of actin rods (33). Mitochondrial dysfunction concomitant with ATP decline (59,60) as well as, though controversial, elevated oxidative stress (49,61) have been described for SMA. These factors contribute to rod formation thereby linking SMN loss to the presence of actin rods (Fig. 8). However, what could be the cause of PLXND1 cytoplasmic fragment binding to actin rods? Interestingly, it has been shown that overexpression of PLXND1’s cytoplasmic domain alone is sufficient to induce cell death likely by translocation into the mitochondrial membrane (62). Since transient sequestration of hypophosphorylated cofilin into rods is a protective mechanism to prevent mitochondrial shuttling, it can be hypothesised that rods function as a sink for the cytoplasmic PLXND1 fragment to prevent mitochondrial translocation. Nonetheless, our results give no hint that PLXND1 cleavage induces rod formation suggesting that both processes are independent from each other.
As mentioned above, we decreased PLXND1-positive rod numbers and corrected altered signal transduction by the use of MP inhibitors. As there is emerging evidence of a role of MPs in neurodegeneration (63), our data suggest that MPs display novel and valuable targets for a combinatorial therapeutic strategy. Such a therapy should be able to rescue defects like altered axonal guidance, malformation of sensory circuits and decreased vascularisation in muscle and spinal cord in SMA.
Materials and Methods
Cell culture, siRNAs, plasmids and compounds
Motoneuron-like NSC34 cells (murine neuroblastoma x spinal cord hybrid cell line (64)) were incubated at 37 °C in a humidified atmosphere with 5% CO2. Cells were maintained in DMEM (Gibco GlutaMAX) supplemented with 5% FCS, 100 U ml−1 penicillin and 0.1 mg streptomycin. Tests for mycoplasma were routinely performed. 24 h after seeding, medium was changed to low serum conditions (1% FCS) to induce differentiation. Cells were then transfected with scrambled control siRNA (siCtrl) or siRNA against murine Smn (siSmn) using Lipofectamine2000 according to the manufacturer’s instructions and allowed to differentiate for three days after which Smn knockdown efficiency was about 90% (18). pBKCMV-VSV-hPLXND1 was a gift of Dr. F. Mann, Aix Marseille Univ, CNRS, IBDM, Marseille, France. siRNA against Plxnd1 was 5’-GCAAGAAAGUAUUGCCAGA-3’ (16). Batimastat (BB-94; Selleckchem) in DMSO was added at 2 µM for the whole time of differentiation and medium was replaced each day. Recombinant SEMA3E (R&D Systems, 3238-S3-025) in PBS was added at 1 nM for 30 min prior to fixation.
Generation of iPSCs and motoneuron differentiation
iPSC induction and motoneuron differentiation were performed as described previously (40,65) with slight modifications. In brief, human fibroblasts obtained from 2 SMA type I patients and 2 healthy matched controls were transduced with a lentiviral vector expressing Oct-4, Klf5, SOX-2, and c-Myc to generate iPSCs (66). A summary of patient and control information can be found in Supplementary Material, Table S1. Control fibroblasts were kindly provided by Dr. Vivi M. Heine (VU University, Amsterdam, The Netherlands). iPSCs were cultured on a confluent layer of irradiated mouse embryonic fibroblasts (MEFs) in human embryonic stem cell (huES) medium, containing DMEM-F12 (Life Technologies), knockout serum replacement (Life Technologies), penicillin/streptomycin (Life Technologies), L-glutamine (Life Technologies), nonessential amino acids (Life Technologies), β-mercaptoethanol (Merck Millipore, Billerica, MA), and 20 ng ml−1 recombinant human fibroblast growth factor-basic (Life Technologies). Colonies of iPSCs were manually picked for further expansion and characterisation. Feeder-free culture of iPSCs was carried out on Geltrex (Life Technologies) and maintained in mTeSR1 medium (STEMCELL Technologies, Vancouver, BC, Canada). Feeder-free cultured iPSCs were passaged enzymatically using Accutase (Innovative Cell Technologies, San Diego, CA.). All cell lines were routinely (every 2 weeks) tested for mycoplasma infections. Characterisation of iPSC lines was performed by immunocytochemistry for pluripotency markers (Stem Light Pluripotency Antibody Kit, BIOKÉ), pluripotency gene expression analysis (qRT-PCR), karyotyping, and spontaneous differentiation assays (Verheijen and Pasterkamp, unpublished).
Motoneurons were differentiated as described previously (67) with some modifications. iPSCs were first detached and seeded into microwells (68,69) in huES medium supplemented with 10 μM ROCK inhibitor Y-27632 (Axon Medchem, Groningen, the Netherlands) to reduce cell death. After 48 h, medium was changed to neural induction medium (NIM) containing DMEM-F12 (Life Technologies), penicillin/streptomycin (Life Technologies), L-glutamine (Life Technologies), nonessential amino acids (Life Technologies), N2 supplement (Life Technologies), and 20% D-glucose (Sigma). For neuralisation of the resulting embryoid bodies (EBs), dual SMAD signalling was inhibited between days 1 and 5 using 10 μM SB431542 (Axon Medchem) and 0.2 μM LDN193189 (Miltenyi Biotec, Bergisch Gladbach, Germany). For the first 4 days, 10 μM Y-27632 (Axon Medchem) was added to inhibit cell death. EBs were flushed out of the microwells using NIM containing 10 μM SB431542 (Axon Medchem), 0.2 μM LDN193189 (Miltenyi Biotec), 1 μM retinoic acid (Sigma), and 10 ng ml−1 brain-derived neurotrophic factor (BDNF, R&D Systems, Minneapolis, MN) and transferred to a non-adherent petri dish (Greiner Bio-One, Monroe, NC). EBs were kept in suspension, and medium was changed every other day using NIM containing 1 μM retinoic acid (Sigma), 1 μM smoothened agonist (Merck Millipore), and 10 ng ml−1 BDNF (R&D Systems). Starting on day 16, medium was changed every other day using Neurobasal differentiation medium containing Neurobasal (Life Technologies), penicillin/streptomycin (Life Technologies), L-glutamine (Life Technologies), nonessential amino acids (Life Technologies), N2 supplement (Life Technologies), B27 minus vitamin A (Life Technologies), and 20% D-glucose (Sigma) supplemented by 1 μM retinoic acid (Sigma), 1 μM smoothened agonist (Merck Millipore), 10 ng ml−1 BDNF (R&D Systems), 10 ng ml−1 glial cell line-derived neurotrophic factor (GDNF, R&D Systems), and 10 ng ml−1 ciliary neurotrophic factor (CNTF, R&D Systems). From days 21 to 31, EBs were dissociated using papain (Worthington Biochemical Corporation, Lakewood, NJ) and DNAse (Worthington Biochemical Corporation). Cells were resuspended in human motoneuron medium containing Neurobasal (Life Technologies), penicillin/streptomycin (Life Technologies), L-glutamine (Life Technologies), nonessential amino acids (Life Technologies), N2 supplement (Life Technologies), and B27 minus vitamin A (Life Technologies) supplemented by 10 ng ml−1 BDNF (R&D Systems), 10 ng/ml GDNF (R&D Systems), and 10 ng ml−1 CNTF (R&D Systems) and plated on poly-D-lysine (PDL)-laminin–coated coverslips at the required density. Coverslips with MNs were co-cultured with primary mouse glia for max. 2 weeks. AraC treatment was used to eliminate progenitor cells from the culture. Motoneuron identity was confirmed by immunocytochemistry for motoneuron markers, including Hb9, Isl-1, and ChAT. For the analyses, four iPSC clones for each line were generated and one clone per line was selected to work with.
Animals
The mouse mutant strain FVB.Cg-Tg(SMN2)2Hung SMN1tm1Hung/J (19) was purchased from the Jackson Laboratory (stock number 005058). To obtain 50% SMA-mice (Smn−/−;SMN2tg/0) and 50% control littermates (Smn+/−;SMN2tg/0), mice were bred and genotyped as described previously (10,70). After decapitation, spinal cord was dissected and immediately frozen in liquid nitrogen or prepared for cryosection. All experimental protocols followed German animal welfare law and were approved by Lower Saxony State Office for Consumer Protection and Food Safety (LAVES, reference number 15/1774).
The Taiwanese severe SMA model was crossed back on C57BL/6 N background and treated with low dose SMN-ASOs by subcutaneous injection of 30 µg at P2 and P3 to produce an intermediate SMA phenotype as described in detail (22). Animal care and all surgical procedures were performed according to the institutional animal care committee guidelines and the German animal welfare laws and approved under the reference numbers 84-02.04.2014.A006 and 84-02.04.2015.A378 of the LANUV state agency of North Rhine-Westphalia.
Quantitative reverse transcriptase (qRT) PCR
Total mRNA from cells or tissue was isolated using the Qiagen RNeasy Plus Kit according to the manufacturer’s instructions. cDNA was synthesized with the SuperScript IV First Strand Synthesis System (Invitrogen) Kit following the Kit’s recommendations. qRT-PCR was performed with the SYBR Green reaction mix (Applied Biosystems) in a StepOnePlus thermocycler (Applied Biosystems) with StepOne V2.3 software. The following primers (5’->3’) were used: Plxnd1 (NM_026376.3) Fwd CAGTACTGCCCTCGGAGATTG and Rev TCGAACGCCATTTCCATAGTC, Hprt1 (NM_013556.2) Fwd TTCCTCATGGACTGATTATGGACA and Rev AGAGGGCCACAATGTGATGG (71) and Ppia (NM_008907.1) Fwd TGCACTGCCAAGACTGAATG and Rev CCATGGCTTCCACAATGTTC (72). Primer specificity was evaluated by melting curve analysis. CT values between technical replicates had to be smaller 0.4 cycles at a threshold of 0.2. CT values were analysed using the 2-ΔCT method (73) and Hprt1 was used as reference gene as known to be stably expressed (18). Normalisation was conducted over the mean of all 2-ΔCT values of the whole group to allow for comparison between time points.
Western blot analysis
Cells were washed with PBS, scraped into RIPA buffer containing protease and phosphatase inhibitors (Roche) and allowed to chill on ice for 20 min (74). Tissue samples were homogenised in RIPA buffer containing protease and phosphatase inhibitors (Roche) and also allowed to chill on ice for 20 min. Afterwards, lysates were sonicated and clarified by centrifugation (21,000 × g, 20 min, 4 °C). Protein concentration was determined with the BCA Assay (Thermo Scientific). To detect PLXND1 with an apparent size of 250 kDa, equal amounts of protein were dissolved in Lämmli buffer and separated by 6% SDS-PAGE. Protein was subsequently blotted onto a nitrocellulose membrane for 3 h at 30 V in the cold. Proteins smaller than 150 kDa were dissolved in Lämmli buffer and separated by 12.5% SDS-PAGE and blotted for 1 h at 120 V. Primary antibodies were: mAB mouse α-actin (BD Bioscience 612656, LOT 26375, 1:4000), mAB rabbit α-cofilin (Cell Signaling #5175 S, LOT 2, 1:1000), pAB rabbit α-PlexinD1 (Abcam ab96313, LOT GR145861-8, 1:1000), mAB mouse α-SMN (BD Bioscience 610647, LOT 4157975, 1:4000) and mAB mouse α-α-tubulin (Santa Cruz sc-32293, LOT D0814, 1:4000). HRP-conjugated secondary antibodies (GE Healthcare, 1:4000) were used for visualisation by chemiluminescence imaging using the Immobilon reagent (Millipore) or the SuperSignal® West Femto Substrate (Thermo Scientific). Densitometric analysis was performed with the LabImage 1 D software (Kapelan, Leipzig, Germany). Relative band intensities were calculated from bands of the same blot. Other targets were analysed on the same blot after stripping of the membrane (62 mM Tris-HCl pH 6.8, 2% SDS and 0.7% β-ME) for 30 min at 50° C). Cropped blots are indicated in the figure legends. Uncropped blots are shown in Supplementary Material, Figure S8.
Isolation of actin rods
Rods were isolated from SMN-depleted NSC34 cells with a modified protocol as described elsewhere (28). In brief, differentiated knockdown cells from a 10 cm dish were incubated with 10 µM nocodazole to destroy interfering microtubules 10 min prior to lysis. All lysis and centrifugation steps were carried out at 4 °C. Cells were washed with PBS, scraped into 350 µl of lysis buffer (20 mM PIPES pH 6.8, 140 mM NaCl and 1 mM EGTA) and disrupted by passaging through a 27 gauge needle. The lysate was pre-cleared by centrifugation at 350 × g for 3 min. Rods in the supernatant were then enriched by centrifugation on a discontinuous OptiPrep gradient (10 and 15% (v/v)) for 10 min at 6,600 × g. 500 µl fractions were collected and analysed for cofilin and actin by Western blot. The actin rod-containing fraction was then diluted 1:2 with lysis buffer and centrifuged for 15 min at 21,000 × g. The pellet was either dissolved in Lämmli buffer, sonicated and analysed for PLXND1 by Western blot or directly used for electron microscopy.
Immunocytochemistry
Cells were washed with PBS and fixed with 4% PFA for 10 min. Afterwards cells were permeabilised with ice-cold methanol for 10 min and extensively washed with PBS. Alternatively, cells were permeabilized with PBS containing 1% Triton X-100 and 4% horse serum for 20 min at room temperature or for 10 min with ice-cold methanol, which is best to maintain rods (28), followed by 20-min Triton X-100 treatment. Primary antibodies in PBS containing 1% horse serum were added for 1 h at room temperature or overnight at 4 °C. Primary antibodies were: mAB mouse α-actin (BD Bioscience 612656, LOT26375, 1:500), goat α-ChAT (Millipore), mAB rabbit α-cofilin (Cell Signaling #5175 S, LOT 2, 1:500), mouse α-Hb9 (Developmental Studies Hybridoma Bank, DSHB), mouse α-Isl-1 (DSHB), pAB goat α-PlexinB1 (Santa Cruz sc-28372, LOT E1314, 1:200), pAB goat α-PlexinD1 (Abcam ab28762, LOT GR41626-16, 1:200), pAB rabbit α-PlexinD1 (Abcam ab96313, LOT GR145861-8, 1:200), pAB goat α-PlexinD1 (Santa Cruz sc-46246, LOT I1307, 1:200), rabbit α-Tuj1 (Sigma) as well as Alexa546-coupled phalloidin (Thermo Scientific A22283, LOT 1351889, 1:200). Alexa-coupled secondary antibodies (Invitrogen, 1:500) in PBS containing 1% horse serum were added for 1 h at room temperature. DAPI in PBS was added for 5 min at room temperature. Cells were mounted in Prolong Gold (Life Technologies).
Immunohistochemistry
Cryo-sections (10 µm thickness) of lumbar spinal cord for staining with pAB goat α-PlexinD1 (Abcam ab28762, LOT GR41626-16, 1:200) or pAB rabbit α-PlexinD1 (Abcam ab93234, LOT GR64899-1, 1:200) were treated with 10 mM citrate buffer pH 6 at 95 °C for 10 min prior to staining. The pAB rabbit α-PlexinD1 (Abcam ab96313, LOT GR145861-8, 1:200), pAB goat α-PlexinD1 (Santa Cruz sc-46244, LOT D2407, 1:200) and pAB goat α-PlexinD1 (Santa Cruz sc-46246, LOT I1307, 1:200) antibodies were used without antigen retrieval. In brief, tissue sections were blocked for 1 h with 3% (w/v) BSA and 0.3% (v/v) Triton X-100 in PBS and incubated overnight at 4 °C with primary antibodies: mAB mouse α-SMI32 (Sternberger Monoclonal Inc. 801701, LOT D14JF02098, 1:1000), mAB mouse α-GFAP (Sigma G 3893, LOT 083M4785, 1:600). Secondary antibodies in PBS containing 1% (w/v) BSA and 0.3% (v/v) Triton X-100 were the same as used for immunocytochemistry and added for 1 h at room temperature. DAPI in PBS (1:1000) was added for 5 min and slices were mounted in Prolong Gold (Life Technologies).
Microscopy and evaluation of images
Epifluorescence images were taken using an Olympus BX60 upright fluorescence microscope equipped with an Olympus XM10 color view camera and Olympus Cell Sense software. Exposure time was equal for each channel in the respective experiment. Background reduction (rolling ball radius: 50 pixels) and brightness adjustment were performed over the whole image using ImageJ software. Parameters were kept equal for all images to enable comparison within replicates. To analyse the percentage of NSC34 cells showing rods, rod-positive cells were counted relative to total cells from ten randomly taken images of each coverslip. Data from technical replicates were averaged and used for statistics for all analyses. Where possible, evaluation of samples was carried out blinded.
Confocal images were taken using a Leica TCS SP2 equipped with an oil immersion objective HCX PL APO BL (63×, numeric aperture 1.4) with Leica acquisition software, a Zeiss LSM 880 equipped with an Zeiss oil immersion objective Plan-Apochromat (63×, numeric aperture 1.4) or a Zeiss META 510 confocal microscope with ZEN software (Zeiss) to create Z stacks. Co-localisation was analysed within the indicated ROIs using the Intensity Correlation Analysis plugin for ImageJ with default parameters after global background reduction (rolling ball method, radius 50 pixels) to determine product of the differences of the mean (PDM) values.
For electron microscopy, hanging drops of centrifuged fractions were incubated on a carbon-coated formvar copper grid (Science Services, München, Germany), washed once with Millipore water and negatively stained with 2% (w/v) uranyl acetate in water. Digital images were documented on a Tecnai G 2 (Fei, Eindhoven, The Netherlands) with an accelerating voltage of 200 kV.
Statistics
Statistical analyses were carried out using the GraphPad Prism 6 software (La Jolla, CA, USA). Numbers of biological replicates and tests performed are stated in the figure legends. Animals were randomly picked from different litters. Results were considered as significant if P < 0.05. For morphological evaluation (quantification of rods and growth cone collapse), at least ten randomly picked images from each coverslip were analysed and results were averaged. For each condition, the experiment was carried out blinded in at least three independent biological replicates.
Supplementary Material
Supplementary Material is available at HMG online.
Acknowledgements
We thank Dr. Andreas Ratzka for providing the Plxnd1 primers and Dr. F. Mann, Aix Marseille Univ, CNRS, IBDM, Marseille, France, for providing plasmids. Furthermore, we thank Sandra Kling, Oliver Harschnitz, Liset Rietman and Lill Eva Johansen for help with generating iPSC lines and motoneuron differentiation.
Conflict of Interest statement. None declared.
Funding
Niedersachsen‐Research Network on Neuroinfectiology (N‐RENNT) of the Ministry of Science and Culture of Lower Saxony, the Initiative SMA, and the Deutsche Muskelstiftung/Philipp & Freunde - SMA Deutschland e.V. to P.C, by a grant from SMA Europe (to N.Hen.), by a grant from Stichting Spieren voor Spieren to W.L.v.d.P., the ALS Stichting (TOTALS to R.J.P) as well as the Deutsche Forschungsgemeinschaft Wi945/14-3, RTG1970 and CMMC C11 to B.W.
